Analyzing a comprehensive database of limited liability manufacturing firms this paper investigates the relation between a firm's financial situation and its conditional expected growth rate. Specifically, using quantile regressions, we obtain a quantitative characterization of this relation for different quantiles of the growth rates distribution. We find that simple location-shift models, as for instance the OLS, provide a poor and potentially misleading representation of the growth-finance relation. Indeed, the vast majority of the explanatory variables considered are associated with modifications in the support of the growth rates distribution (scale-effect), even when the relation of the same variables with the expected growth is negligible. Moreover, we show that financial conditions impact differently on the growth dynamics of young and old firms. Finally, our investigations reveal that the results obtained with quantile regressions appear robust with respect to possible mispecifications of the empirical model.